racial and ethnic preferences and consequences at the … · • black applicants to umsm are given...

Racial and EthnicPreferences and Consequencesat the University of MarylandSchool of Medicine

By Robert Lerner, Ph.D. andAlthea K. Nagai, Ph.D.

EMBARGOEDUNTIL

April 3, 2001

Prepared for theCenter for Equal Opportunity

Linda Chavez, President815 15th Street, NW, Suite 928Washington, DC 20005Phone: 202-639-0803Fax: 202-639-0827http://www.ceousa.org

http://www.ceousa.org/

ii

Table of ContentsExecutive Summary 1Acknowledgments 3Introduction 4Medical Schools’ Background 6Methodology 10

University of Maryland School of Medicine Preferences 12

Applicants, Admittees, and Enrollees1996 / Page 12Applicants, Admittees, and Enrollees1999 / Page 13Differences in MCAT Scores1996 and 1999 / Page 13Differences in Science GPAs1996 and 1999 / Page 15Rejectees vs. Admittees / Page 17Logistic Regression Analysis and Odds Ratios / Page 17Relative Odds of Admission to UMSM / 0Probabilities of Admission to UMSM / Page 21

Consequences of Racial and Ethnic Preferences at UMSM 26

Early Help for URMs / Page 26Analyzing Performance in Medical School / Page 28Differences in Medical School GPAs the First Two Years / Page 28Differences in Medical School GPAs the Third and Fourth Years / Page 30Differences in Step 1 Scores / Page 30Differences in Step 2 Scores / Page 32Relationship between MCATs and Subsequent Performance in

Medical School / Page 33Differences in Graduation Rates / Page 35Other Consequences of Racial and Ethnic Preferences / Page 36

Appendix�Multiple Logistic Regression Equations, 1996-1999 38

Figures

Figure 1. Weighted Total MCAT Scores, 1996 UMSM Admittees / Page 14Figure 2. Weighted Total MCAT Scores, 1999 UMSM Admittees / Page 15Figure 3. Science GPAs, 1996 UMSM Admittees / Page 16Figure 4. Science GPAs, 1999 UMSM Admittees / Page 16

iii

Figure 5. Minority-to-White Applicant Odds Ratios, UMSM 1996-1999 / Page 20Figure 6. 1996 UMSM Probabilities of Admission / Page 22Figure 7. 1999 UMSM Probabilities of Admission / Page 24Figure 8. 1st & 2nd Year GPAs, UMSM Enrollees / Page 29Figure 9. 3rd and 4th Year GPAs, UMSM Enrollees / 0Figure 10. USMLE Step 1 Scores, UMSM Enrollees / Page 31Figure 11. USMLE Step 2 Scores, UMSM Enrollees / Page 32

1

Executive SummaryPreferences

• Black applicants to UMSM are given a massive degree of preference over theirwhite, Hispanic, and Asian counterparts.

• Hispanics, Asians, and whites are admitted with roughly the same MCAT scores.Hispanic and white science GPAs are roughly the same, while science GPAs forAsian admittees are slightly lower.

• UMSM generally admits blacks with much lower test scores and science GPAs ascompared with whites, Hispanics, and Asians. In 1996 and 1999, 75 percent ofblacks were admitted with test scores lower than roughly 75 percent of all whiteand Hispanic admittees. In those years, the median science GPA for blackadmittees was lower than the science GPA for 75 percent of Hispanics and whitesadmitted by UMSM. Seventy-five percent of blacks were admitted by UMSMwith lower scores than half the Asians admitted in 1996, while 75 percent ofblacks were admitted with lower scores than 75 percent of Asians admitted in1999.

• The relative odds of admission of a black over a white applicant were 61.5 to 1 in1996, 35.9 to 1 in 1997, 40.7 to 1 in 1998, and 20.6 to 1 in 1999.

• The odds ratios translate into massive preferences favoring blacks over othergroups. For example, in 1996, the probability of admission for a black applicantwith a total MCAT score of 45 and a science GPA of 3.5 was 97 percent—roughly three times the probability of a similarly qualified white, Asian, orHispanic. The probability of admission for a white applicant with the samecredentials was 33 percent; for an Asian, 28 percent; and for a Hispanic, 37percent. For 1999, the probability of admission of a black applicant with a totalMCAT score of 45 and a science GPA of 3.5 was 90 percent, as compared to 31percent for a similar white applicant, 23 percent for an Asian applicant, and 53percent for a Hispanic applicant. These statistics control for sex, residency, andwhether the person had a parent graduating from UMSM.

Consequences• Black enrollees generally have much greater difficulty in medical school than do

whites, Asians, and Hispanics, despite UMSM’s massive program of academicintervention and remediation specifically for “underrepresented minorities.”

• The median medical school GPA in the first two years was 2.50 for blacks, 3.00for Hispanics, and 3.17 for Asians and for whites.

2

• The median medical school GPA for the third and fourth years is 3.29 for blacks,3.50 for Hispanics, 3.50 for Asians, and 3.38 for whites.

• UMSM black enrollees perform considerably worse on the medical licensingexams than do their Hispanic, Asian, and white counterparts, again despiteUMSM’s academic intervention and remediation for underrepresented minorities.

• More than a quarter of the black enrollees (7 out of 27) failed “Step 1” of themedical licensing exam on their first try. Two whites, one Hispanic, and noAsians failed. The median Step 1 score for black enrollees was roughly the sameas that for Hispanics, but lower than that for 75 percent of Asian and whiteenrollees.

• About a quarter of the black enrollees (4 out of 15) taking “Step 2” of the medicallicensing exam failed it on their first try. No student from another group failed.

• The four-year graduation rate for blacks was 68 percent. Blacks graduated at ahigher rate than do Asians (63 percent), but at a much lower rate as comparedwith whites (82 percent). Hispanic graduation rates are not reported.

3

AcknowledgmentsWe thank Linda Chavez and her staff at the Center for Equal Opportunity for

giving us the chance to work on another major study of racial and ethnic preferences.Several individuals deserve special mention. Hugh Joseph Beard, Jr., obtained the

basic data. John Montgomery made available his unsealed data and documents. Bothworked to have the materials from Farmer v. Ramsay unsealed for public view.

Roger Clegg offered comments and criticisms and tirelessly edited this report.Sally Satel and John Montgomery read drafts of this report and offered insights,

comments, and criticisms.The statistical analyses, conclusions, and any mistakes are our own.

4

IntroductionFor more than thirty years, racial and ethnic preferences have played a key role in

how admissions officers at the nation’s public and private colleges and universities havechosen their undergraduate classes. A system of racial and ethnic preferences inadmissions operates by establishing different standards of admission for individuals basedupon their racial or ethnic background, with some students held to a higher standard andothers admitted under a lower standard. Early in the last century, some colleges anduniversities denied admissions to Jews, blacks, women, and members of other groupseven when their grades, test scores, and other measures of academic achievementsurpassed those of white males who were offered an opportunity to enroll. The passage ofcivil rights legislation in the 1960s made this kind of discrimination illegal.

Since then, however, many colleges and universities have created “affirmativeaction” programs meant to boost the enrollment of students from groups whosebackgrounds previously had excluded them from pursuing higher education, includingpost-baccalaureate education—especially blacks and, to a lesser extent, Hispanics—bygranting them preferences during the admissions process. These policies, when theirexistence was made public, immediately became controversial, and they remain so today.Defenders of racial and ethnic preference policies claim that these policies are notdiscriminatory and help administrators choose between equally or almost equallyqualified students, giving a slight edge to applicants who likely have faced discriminationor have come from disadvantaged backgrounds. Critics of preferences say that thesepolicies are no better than the discriminatory ones they replaced and that, in any event, theadvantages they confer upon certain applicants are much greater than supporters arewilling to admit.

Roughly fifteen years ago, sociologist William Beer lamented the dearth ofempirical studies of preference programs and their consequences.1 The situation hasimproved somewhat, but the extent, operation, and consequences of racial and ethnicpreferences in higher education remain one of the nation’s better-kept secrets. There havebeen only grudging concessions that preferences have been used in admissions—or as theauthors of The Shape of the River have put it, that admissions have been “raciallysensitive.”2

In the last few years, public colleges and universities have seen their ability to useracial and ethnic preferences increasingly restricted. The enactment of California’sProposition 209 (also known as the California Civil Rights Initiative) forbidsdiscrimination against or granting special treatment to any applicant on the bases of race,ethnicity, or sex in the public programs of the country’s largest state. A similar ballot

1 William Beer, “Resolute Ignorance: Social Science and Affirmative Action,” Society (May/June 1987):63-69.2 See Robert Klitgaard, Choosing Elites (New York: Basic Books, 1985); Thomas Kane, “Racial and EthnicPreferences in College Admissions,” in Christopher Jencks and Meredith Phillips, eds., The Black-WhiteTest Score Gap (Washington, D.C.: The Brookings Institution, 1998): 431-56; and William G. Bowen andDerek Bok, The Shape of the River (Princeton: Princeton University Press, 1998).

5

initiative in Washington state, Initiative 200, was approved by a large majority of votersin 1998. Court decisions have also limited the use of such preferences. Florida, Texas,and California have all created policies that guarantee admission to the state universitysystem to the top graduates of their respective state’s high schools regardless of race orethnicity.

The studies published by the Center for Equal Opportunity have been the onlystudies, to our knowledge, to uncover and systematically document the discrimination inundergraduate admissions among America’s public colleges and universities. EarlierCEO studies have focused on the public colleges and universities of Colorado, Maryland,Michigan, Minnesota, North Carolina, and Virginia, the University of Washington andWashington State University, the U.S. Military Academy and U.S. Naval Academy, aswell as the branches of the University of California at Berkeley, Irvine, and San Diego.These reports, summarized and expanded upon in the recently issued CEO monograph,Pervasive Preferences: Racial and Ethnic Discrimination in Undergraduate Admissionsacross the Nation, have shown that blacks and, to a lesser extent, Hispanics receivesubstantial degrees of preference in public undergraduate admissions across the country.

The focus now shifts to professional schools and to subsequent performance inprofessional schools as a function of racial and ethnic preferences. This report on theUniversity of Maryland School of Medicine (UMSM) is the first in a series that willanalyze the extent of racial and ethnic preferences in law and medical school admissions.Additionally, this and subsequent CEO reports will investigate the consequences of racialand ethnic preferences on subsequent performance once students are enrolled.

Previous CEO studies of preferences in public undergraduate institutions of highereducation have obtained some aggregate data on graduation rates for racial and ethnicgroups. These have shown that blacks and Hispanics are less likely to graduate frominstitutions giving them admission preferences than are their white and Asiancounterparts. Aggregate graduation rates are limited measures of academic performance,however, because they reflect at best a minimum standard of academic achievement.

This study of UMSM examines both the use of racial and ethnic preferences inadmissions and the medical school performance of those that subsequently enrolled. Thereport examines the efforts extended by UMSM to keep these students in medical school.It also presents statistical evidence of how well or poorly they performed in class and onthe critical medical licensing exams.

6

Medical Schools’ BackgroundIncreasing underrepresented minority (URM) admissions to medical schools has

been a major project of the academic medical establishment for many years.3 The lateBernard D. Davis, Emeritus Professor at Harvard Medical School, recounts his firsthandexperience of how Harvard began to award racial and ethnic preferences in admissions tomedical school. Davis pointed out that, after the murder of Dr. Martin Luther King, Jr.,the Harvard Medical School decided to admit a substantial number of black students whootherwise lacked the requisite qualifications. Not surprisingly, they performed poorly.Rather than abandoning preferences, Harvard Medical School chose to lower classroomstandards. The decision was made with no open faculty debate. Departments wererequired to allow failing students to retake exams until everyone passed, letter gradeswere replaced by a pass/incomplete system (and, once a student had passed, he or sheretained no trace of the incompletes), the number of required courses was reduced whilethe number of electives was substantially increased, passing scores on the nationallicensing exams were lowered, and one minority student was even allowed to graduatefrom Harvard after having failed the required medical licensing exam five times.4

Davis’s experiences appear to be standard fare. The American Association ofMedical Colleges (AAMC) and the American Medical Association (AMA) have made aconcerted effort since the 1970s to increase the number of underrepresented minorityphysicians in America. The AAMC has collected statistics on racial and ethnic groupsapplying, enrolling, and completing medical school since 1960. Comparing thesepercentages to the percentages of groups in the general population, the medicalestablishment has decided that certain groups—Hispanics, blacks, and NativeAmericans5—are underrepresented when compared with their percentage of the U.S.population. Underrepresented minority enrollment was 10.3 percent in 1992, but thesegroups made up 22.1 percent of the U.S. population.6

The AAMC is strongly committed to the goal of proportional representation andencouraged medical schools across the country to graduate 3000 URM doctors by the 3 See Sally Satel, PC, M.D.: How Political Correctness Is Corrupting Medicine (New York: Basic Books,2000) on the work of the AAMC and others regarding racial preferences in medical education and beyond,as part of the general politicization of health-care groups. The general summary of the political activities ofthese health-care groups comes from her book.4 Bernard D. Davis, “Affirmative Action and Veritas at Harvard Medical School,” Storm over Biology(Buffalo, New York: Prometheus Books, 1986): 169-191.5 The AAMC classification system appears standardless and arbitrary. Before 1993, “Native Americans”included only Native Alaskans and American Indians, while Native Hawaiians were classified asAsian/Pacific Islanders. Since 1993, “Native Americans” has included Native Hawaiians as well as NativeAlaskans and American Indians. Other Americans of Pacific Island descent (e.g., Samoan) are stillclassified as “Asian/Pacific Islander” and thus are not URMs. Association of American Medical Colleges,AAMC Data Book: Statistical Information Related to Medical Schools and Teaching Hospitals, January1999 (Washington, D.C.: AAMC): 15.6 Committee on Increasing Minority Participation in the Health Professions, Institute of Medicine,Balancing the Scales of Opportunity: Ensuring Racial and Ethnic Diversity in the Health Professions(Washington, D.C.: National Academy Press, 1994): 1.

7

year 2000. Politically, the AAMC and other health-care groups have worked activelyagainst attempts to dismantle racial and ethnic preference policies. They came together toform Health Professionals for Diversity in 1996 and worked against passage ofProposition 209 in California. By 1998, the coalition was made up of 51 health-careinterest groups, and actively (but unsuccessfully) campaigned against the passage ofInitiative 200 in Washington state. The AAMC went so far as to run full-page ads in localWashington newspapers, warning voters that passage of Initiative 200 would depriveminorities of medical care. After the passage of Initiative 200, the National MedicalAssociation (NMA), the professional association of black physicians, canceled its 2001convention in Seattle (despite a majority in Seattle voting against Initiative 200).

The University of Maryland School of Medicine has been at the forefront at theeffort to increase the number of minority medical students. UMSM, according to theOffice of Admissions, would “take a lead role in the activities of the Association ofMedical Colleges’ Project 3000 by 2000.”7 As part of its mission, “the School ofMedicine will become recognized for the rich diversity of its student body, matriculatingeach year increasing numbers of women and underrepresented minorities into the MD andMD/PhD Programs . . . .”8 UMSM also states, “Increased diversity among the studentsand faculty at the UMSM has been a high priority of the school for more than 25 yearsand has been enhanced by Dean Donald E. Wilson, M.D., the first African-Americandean of a predominantly non-minority school of medicine in the United States, andadministration of the Campus.”9

UMSM has retrained its admission committee members to pay attention to “non-cognitive variables” for the sole purpose of increasing the number of underrepresentedminorities in medical school. 10 UMSM has sent admission committee members toparticipate in the AAMC’s Simulated Minority Admissions Exercises (SMAE), whichlater became the AAMC’s Expanded Minority Admissions Exercises (EMAE). Theseexercises are conducted by AAMC so admission committee members can learn to look atnonacademic variables and place less weight on academic criteria when selectingminorities for medical school. AAMC does not suggest looking at these noncognitivevariables when assessing whites and Asians.

In 1996, UMSM explicitly stated its intention to use the Simulated MinorityAdmissions Exercises for the purpose of increasing its number of black, Hispanic, and

7 Functional Statement/Office of Admissions, n.d., p. 4.8 Mission Statement/ Office of Admissions, n.d., p. 1-2.9 University of Maryland School of Medicine, Training Grant Application, Health Careers OpportunityProgram, 93.822, 1996, p. 16.10 Research fails to support the AAMC’s contention that noncognitive variables are important medicalschool considerations. For example, Webb et al., 1997, is erroneously cited as evidence that noncognitivevariables are important in predicting medical school performance. This study of two medical schools, A andB, actually shows that noncognitive variables have little or no capacity to predict success in medical schooland beyond. The findings actually show that academic factors at School A were considerably moreimportant in predicting success; noncognitive variables for medical students at School A were barelystatistically significant. At the least, academic factors predicted three times better than noncognitivevariables; at the most, academic factors predicted nine times better than noncognitive factors. At School B,noncognitive factors were not statistically significant at all; they predicted nothing.

8

Native American medical students.11 UMSM trains its admission committee membersthrough these exercises to look for noncognitive factors, “to be used when evaluating aminority applicant’s application.”12 Additionally, underrepresented minority applicantsmeet the Associate Dean for Admissions, have lunch with current medical students,receive a tour of the medical school, and/or meet the director of recruitment orrecruitment coordinator prior to their interviews.13

UMSM’s commitment to increase minority enrollment is also reflected in itsreports that track minority enrollment and retention. In these documents, UMSM and theother schools that make up the University of Maryland at Baltimore (e.g., theundergraduate programs, the law school, and the schools of public health, pharmacy,nursing, and social work) provide yearly reports on such matters as each school’sminority enrollment goals, the actual number of minorities enrolled, support servicesprovided minorities, their retention rates, and their eventual graduation rates.

According to the University of Maryland at Baltimore’s 1994 Minority EnrollmentReport, 12 percent of medical students were black in 1993. At this time, UMSM set itsfive-year enrollment goal for blacks at 14 percent.14 For Hispanics, who made up 2percent of students at the medical school in 1993, UMSM set a projected goal for the1994-through-1998 period at 3 percent.15

In its report the following year, UMSM stated that blacks made up 13 percent ofmedical students in 1994, and it set the projected goal for 1998 at 14 percent.16 Hispanicsmade up 2 percent of medical students in 1994, and the projected 1998 goal was set at 3percent.17

Having reached its black enrollment goal in 1996, UMSM then raised its goal to18 percent for 1997 and subsequent years. 18 According to the 1996 report, blacks madeup 15 percent of medical students in 1995, which was one percent higher than theenrollment goal laid out in previous years. Hispanics made up 3 percent of medicalstudents in 1995, and the projected goal for 1998 was also 3 percent.19

The AAMC recognized UMSM’s success in averaging a minority enrollment of15 percent in the 1990s.20 UMSM was so successful in enrolling blacks, Hispanics, andAmerican Indians that the dean of UMSM became the first recipient of AAMC’s HerbertW. Nickens Award for Diversity. The award goes to an individual “who has madeoutstanding contributions to promoting justice in medical education and health care.” Thepresident of the AAMC commended Dean Wilson for his “dedication to the principles ofdiversity and equity in health care.”

11 University of Maryland at Baltimore, Minority Achievement Report, July 1996, p. 85.12 Ibid.13 Ibid.14 University of Maryland at Baltimore, Minority Achievement Report, July 1994, p. 3.15 Ibid., p. 5.16 University of Maryland at Baltimore, Minority Achievement Report, July 1995, p. 3.17 Ibid., p. 5.18 University of Maryland at Baltimore, Minority Achievement Report, July 1996, p. 3.19 Ibid., p. 5.20 Press Release, “AAMC Names University of Maryland Dean First Recipient of Nickens Award forDiversity,” October 28, 2000, <www.aamc.org/newsroom>.

9

Clearly UMSM is on the cutting edge of minority admission and enrollment.UMSM is also the defendant in a pending reverse discrimination suit, wherein a whiteapplicant, Robert Farmer, is claiming that UMSM discriminated against him because ofhis race when it turned him down for medical school. The medical school denies thecharge.21

21 Farmer v. Ramsay, Civil No.: L98-1585 (U.S. District Court for the District of Maryland). Robert Lerner,one of the authors of this report, is an expert witness for the plaintiff in this case.

10

MethodologyJust as high school seniors seeking college admissions take the SATs or the

ACTs, prospective medical school students must take the Medical College AdmissionTest (MCAT). The MCAT is a standardized test made up of three multiple-choicesubtests and, since 1993, a writing section. The three subsections are the verbal reasoningsection, the physical sciences section, and the biological sciences section.22 MCATscience subtests are achievement tests, not aptitude tests. They measure knowledge, notintelligence.

The physical sciences, biological sciences, and verbal reasoning subtests are givensubscores, each ranging from a low of 1 to a high of 15. In addition, the writing sample isgiven a letter grade, ranging from J to T.

CEO obtained the data on individual applicants’ admission status (accept orreject), matriculation status (enroll or not), racial or ethnic group membership, sex, stateof residency, whether a parent had graduated from UMSM, UMSM’s weighted totalMCAT score, individual MCAT subscores, and undergraduate science, nonscience, andoverall college GPAs. The weighted total MCAT scores were provided by UMSM as partof the data obtained by CEO through a freedom-of-information request. As noted above,the MCAT is made up of four subtests: verbal reasoning, physical science, biologicalscience, and writing. USMS’s weighted total score is obtained by doubling the writingscore and adding that number to the physical science, biological science, and verbalreasoning subscores.

While data were obtained for UMSM from 1993 through 1999, the focus below isfor the most part on admissions data from 1996 and 1999. Additional statistical analyseswere performed on data from 1996 through 1999 and are included in relevant sections onodds ratios. We omit from our data analyses those cases for which ethnicity is listed as“other,” “missing,” or “unknown.” We also omit Native Americans because of their smallnumbers in this context. Lastly, we omit cases with missing academic data.

In addition, we obtained data on the subsequent performance of those whoenrolled at UMSM in the fall of 1996. We have data on their medical school grades,which is reported in the form of their GPAs for their first and second years, and theirGPAs for their third and fourth years in medical school. We have information on whetheror not these individuals graduated from medical school or not. Finally, we have theirscores on the first two parts of the United States Medical Licensing Examination, theUSMLE Step 1 (up through the third try) and the USMLE Step 2 (first try only).

We do not report group means for test scores or GPAs. Using group means placesgreater weight on extreme values than is warranted. A few unusually high or low scorescan have a substantial effect on the value of the mean. Standard deviations, which arebased on squared deviations from the mean, are even less useful for describing the spread

22 Association of American Medical Colleges, 1998, MCAT Interpretive Manual (Washington, D.C.:Association of American Medical Colleges): 1-5.

11

of cases for asymmetrical, badly skewed distributions. This is because standard deviationsreflect the mathematical square of these extreme values.

The median, however, and related statistics are far less affected by the values ofextreme cases. The median, or the score at the 50th percentile, represents the middle of thedistribution. Fifty percent of all students have greater scores, and 50 percent have lowerscores.

We also report scores at the 25th and 75th percentile, again to deal with theproblem of extreme cases. While the median represents the middle of the distribution, the25th and 75th percentile scores taken together represent the actual spread of scores. Forexample, a 3.25 GPA at the 25th percentile means that 25 percent of GPAs were below3.25, while 75 percent of GPAs were above it. A GPA of 3.90 means that 75 percent ofscores were below 3.90, while 25 percent were above it.

12

University of Maryland School ofMedicine PreferencesApplicants, Admittees, and Enrollees�1996

3,744 individuals applied for admission to the University of Maryland School ofMedicine in 1996.23 831 were residents of the state of Maryland. 2,913 were nonresidents.UMSM admitted 278 (7 percent of applicants). Twenty-five percent of in-state applicantswere admitted, as were 2 percent of out-of-state applicants; 133 accepted applicantsenrolled. A majority of applicants, admittees, and enrollees was white.

UMSM applicants, 1996

• 12 percent black• 4 percent Hispanic• 28 percent Asian• 55 percent white

UMSM admittees, 1996


UMSM enrollees, 1996


UMSM overall admission rates, 1996

• 12 percent of black applicants• 6 percent of Hispanic applicants• 5 percent of Asian applicants• 7 percent of white applicants

23 Applicants listed as “No response,” “American Indian,” and “Alaskan Native” were dropped from theanalysis.

13

Applicants, Admittees, and Enrollees—1999 2,563 individuals applied for admission to the University of Maryland School ofMedicine in 1999.24 596 were residents of the state of Maryland. 1,967 were nonresidents.Of these, 262 (10 percent of applicants) were admitted—33 percent of residents and 3percent of nonresidents—and 130 enrolled. A majority of applicants, admittees, andenrollees was white.

UMSM applicants, 1999


UMSM admittees, 1999


UMSM enrollees, 1999


UMSM overall admission rates, 1999

• 11 percent of black applicants• 12 percent of Hispanic applicants• 8 percent of Asian applicants• 11 percent of white applicants

Differences in MCAT Scores—1996 and 1999The distribution of MCAT scores by groups was roughly the same in 1996 and in

1999. The overwhelming majority of blacks admitted by UMSM has substantially lowerscores than most Hispanics, Asians, and whites admitted.

24 Ibid.

14

1996. Figure 1 shows the weighted total MCAT score as used by UMSM for 1996admittees.25 They are displayed for each racial and ethnic group. In 1996, Hispanicadmittees had the highest MCAT scores of the four groups at all three percentiles,although the differences at all three levels between Hispanic and white admittees is, atmost, two points.

Figure 1

Weighted Total MCAT Scores, 1996 UMSM Admittees(25th, 50th, and 75th percentiles)

34

40

42

38

47

42

47

42

4950

44

51

25

35

45

55

Blacks (n=54) Hispanics (n=9) Asians (n=63) Whites (n=151)

In 1996, black admittees had substantially lower MCAT scores than the otherthree groups. There is a nine-point gap in median scores between black and Hispanicadmittees, a four-point gap between blacks and Asians, and a nine-point gap betweenblacks and whites.

The total MCAT score for 1996 black admittees at the 75th percentile is 42. This islower than the MCAT score at the 25th percentile of Hispanic admittees. It is the same asthe 25th percentile score of white admittees, and only two points higher than the MCATscore of Asian admittees at the 25th percentile. In other words, 75 percent of all blacksadmitted to UMSM in 1996 had lower MCAT scores than roughly 75 percent of allHispanic, Asian, and white admittees.

1999. A similar pattern is found in 1999 (see Figure 2). White, Hispanic, andAsian total MCAT scores are roughly the same, but black total MCAT scores aresignificantly lower.

25 See the previous section, “Methodology,” for a more detailed discussion on the creation of the weightedtotal MCAT score.

15

Figure 2

Weighted Total MCAT Scores, 1999 UMSM Admittees(25th, 50th, and 75th percentiles)

36

45 45

40

4847

48

44

5152

45

51

25

35

45

55


The median MCAT score for black admittees is eight points lower than themedian score for Hispanic and white admittees, and is seven points lower than the medianscore for Asian admittees.

Moreover, the total MCAT score for black admittees in 1999 is lower than thetotal MCAT score at the 25th percentile for Hispanic, Asian, and white admittees. TheMCAT score for black admittees at the 75th percentile in 1999 is 44, while the MCATscore for Hispanic, Asian, and white admittees at the 25th percentile is 45. Thus, 75percent of all black admittees in 1999 had lower MCAT scores than 75 percent of allHispanics, Asians, and whites admitted in 1999.

Differences in Science GPAs—1996 and 1999There is more overlap in science grades among blacks, Hispanics, Asians, and

whites, although there are gaps here as well. Hispanic and white science GPAs aregenerally higher than black and Asian science GPAs in both years. In general, blackswere admitted with lower science grades as compared to Hispanics, Asians, and whites.

1996. Figure 3 displays undergraduate science GPAs for each group in 1996 bythe 25th, 50th, and 75th percentiles.

16

Figure 3

Science GPAs, 1996 UMSM Admittees(25th, 50th, and 75th percentiles)

3.37

3.49

3.25

3.72

3.43

3.67

3.55

3.843.81

3.00

3.51

3.88

3.00

3.20

3.40

3.60

3.80

4.00


The median Hispanic and white science GPAs are higher than the GPAs forblacks and Asians. The median Asian science GPA is roughly two-tenths of a grade-pointhigher than the black science GPA. Both groups’ median GPAs are lower than the GPAsfor Hispanics and whites at the 25th percentile. This means that roughly half the blacksand Asians admitted to UMSM had lower college grades than 75 percent of Hispanic andwhite admittees.

Figure 4

Science GPAs, 1999 UMSM Admittees(25th, 50th, and 75th percentiles)

3.43

3.503.45

3.72

3.653.69

3.65

3.77

3.87

3.28

3.54

3.90

3.00

3.20

3.40

3.60

3.80

4.00


17

1999. Figure 4 presents science GPA data for 1999. Hispanic and white GPAs areroughly the same, while the GPAs of Asians in 1999 are somewhat lower and blackGPAs are lower still. The median GPA of black admittees is 3.45, which is lower than theGPA at the 25th percentile for Hispanics, Asians, and whites, meaning that half the blackswere admitted with lower science grades compared to roughly 75 percent of all Hispanic,Asian, and white admittees.

Rejectees vs. AdmitteesIn this section, where we compare the academic qualifications of rejectees versus

admittees, we will focus on Maryland residents because so few nonresidents wereadmitted. Only 2 percent of nonresident applicants were admitted to UMSM versus 25percent of Maryland residents.

1996. Among Maryland residents in 1996, UMSM rejected 165 Asians, 104blacks, 34 Hispanics, and 319 whites. Of these, 44 Asians, 6 Hispanics, and 81 whiteswere rejected despite equal or higher science grades than the median GPA of blackadmittees. Similarly, 92 Asians, 14 Hispanics, and 186 whites were rejected despitehaving the same or higher total MCAT scores than the average black admittee. Finally,UMSM rejected 43 Asians, 6 Hispanics, and 86 whites with better grades and higher testscores than the median college GPA and total MCAT scores of black admittees.

1999. UMSM rejected 78 black, 124 Asian, 7 Hispanic, and 189 white Marylandresidents who applied to the medical school in 1999. Of these, 60 Asians, 2 Hispanics,and 104 whites were rejected despite having equal or higher science grades than theaverage black admittee. Similarly, 63 Asians, 4 Hispanics, and 109 whites with equal orhigher MCAT test scores than the black median were rejected. Finally, 19 Asians, 2Hispanics, and 36 whites were rejected despite having the same or higher GPAs andMCAT totals as the average black admittee.

Logistic Regression Analysis and Odds RatiosAdmitting students based on racial and ethnic preferences results in schools

accepting preferred minorities with lower test scores and grades as compared to whitestudents at the same school. Admission officers essentially reach down into the applicantpool and pull up certain students, a practice that necessarily results in at least some whiteswith better credentials than preferred minority admittees being rejected from the sameschools, despite their superior qualifications.

Although the data presented thus far provide substantial evidence of racial andethnic preferences at UMSM, it is possible to make the case even stronger andconsiderably more precise. The most powerful means of assessing the degree of racial andethnic preference in admissions is to develop statistical models that predict the probabilityof admission at a school for members of the different ethnic and racial groups, holdingconstant their qualifications. This is done by computing a multiple logistic regressionequation that predicts admission decisions by race and ethnicity and that includes MCATscores and science grades, among others, as statistical control variables.

We use multiple logistic regression analysis as our statistical technique because ofthe nature of the data provided. One way of conventionally expressing a relationship

18

between the independent and dependent variable is by using correlation coefficients. Anegative correlation coefficient of -1.0 signifies a perfect negative relationship betweenthe independent (predictor) variable and the dependent (or outcome) variable, whereby anincrease in the value of the independent variable yields a decrease in the value of thedependent variable. A positive correlation coefficient of 1.0 signifies a perfect positiverelationship between the two variables; as the independent variable increases, so does thedependent variable. Strictly speaking, however, we cannot use correlations to analyzeadmissions data because correlations and standard multiple regression analysis require adependent variable that is non-binary in form. In the case of an applicant’s admissionstatus, the dependent variable (individual admission status) is binary in form—rejectversus admit. To get around this binary-variable problem, we rely on multiple logisticregression equations and their corresponding odds ratios.

The odds ratio is somewhat like a correlation coefficient, except instead ofvarying from 1.0 to –1.0, it varies between zero and infinity. An odds ratio of 1.0 to 1means that the odds of admissions for the two groups are equal. It is equivalent to acorrelation of zero. An odds ratio greater than 1.0 to 1 means that the odds of members ofGroup A being admitted are greater than those for members of Group B, in precisely theamount calculated. An odds ratio of less than 1.0 to 1 means the members of Group A areless likely to be admitted than those in Group B. The former is similar to a positivecorrelation, the latter similar to a negative correlation.

The statistical technique of multiple logistic regression allows us to presentadmissions data in terms of the relative odds of those in Group A being admittedcompared to Group B while simultaneously controlling for a host of other possiblyconfounding variables. The value of the odds ratio is that it provides a relatively directsummary measure of the degree of racial or ethnic preference given in the admissionsprocess for a particular school.

Logistic regression equations predicting the likelihood of admissions werecomputed for the 1996 and the 1999 UMSM applicant pools, controlling for total MCATscores, science grades, alumni, sex, and in-state residency. We were able to derive theodds of admission from these equations for each minority group relative to that of whites,while simultaneously controlling for the effects of these other variables.26

Logistic regression analysis also allows us to test for statistical significance.Statistical calculations always include what is called a p-value. When results are deemedto be statistically significant, this means that the calculated p-value is less than some pre-determined cut-off level of significance. The level of significance conventionally isreported in the form of “p < .05.” This value means that, with these data, there is aprobability equal to or less than 5 percent that the difference found between one groupand another (e.g., blacks versus whites, Hispanics versus whites, or Asians versus whites,since minority groups are being compared to whites) is due to chance. It is a conventionin statistical studies to use the 0.05 value. In more stringent analyses, 0.01 (one in 100), oroccasionally, 0.001 (one in 1,000) can be used as the cut-off. Any p value greater than0.05 (or the more stringent 0.01) is rejected, and the results are said to be nonsignificant.

26 For a discussion of logistic regression and a more complete discussion of odds ratios, see Alan Agresti,Introduction to Categorical Data Analysis (New York: John Wiley and Sons, 1996).

19

A difference that is statistically significant has very little chance of being the result ofchance—that is, a statistical fluke.

In the next section, we discuss odds ratios derived from comparing blacks towhites, Hispanics to whites, and Asians to whites at UMSM. Statistically significantresults are also noted. The size of the odds ratio reflects the strength of the associationbetween racial or ethnic preference and admission status. An odds ratio equal to or greaterthan 3.0 to 1 is commonly thought to reflect a strong relationship; an odds ratio of about2.0 to 1 reflects a moderate association, while a relative odds ratio of 1.5 or less to 1indicates a weak relationship. Of course, an odds ratio of 1.0 to 1 indicates norelationship.27 Finally, a very strong relationship might be taken to be the roughequivalent of the relative odds of smokers versus nonsmokers dying from lung cancer,which in one well-known study is calculated as 14 to 1.28

27 See David E. Lilienfeld and Paul D. Stolley, Foundations of Epidemiology, 3rd edition (New York:Oxford University Press, 1994): 200-202.28 Taken from a 20-year longitudinal study of British male physicians by R. Doll and R. Peto, as quoted inAgresti, Introduction to Categorical Data Analysis, p. 47.

20

Relative Odds of Admission to UMSM

Figure 5

Minority-to-White Applicant Odds Ratios, UMSM 1996-1999

61.52

35.93

40.74

20.63

1.19 1.93 2.520.680.57 0.620.380.77

0

10

20

30

40

50

60

70

1996 1997 1998 1999

Odds

Ratio

s

Black to White Hispanic to White Asian to White

Group 1996 1997 1998 1999Black toWhite

61.52* 35.93* 40.74* 20.63*

Hispanicto White

1.19 0.57 1.93 2.52

Asian toWhite

0.77 0.38** 0.62# 0.68

*Statistically significant at p < .0001

**Statistically significant at p < .001#Statistically significant at p < .05

1996. As displayed in Figure 5, UMSM awards an extremely large degree ofpreference to blacks�but not to Hispanics and Asians�over whites. Controlling forMCAT scores, science grades, sex, residency, and alumni status, the relative odds of ablack applicant being admitted over a white applicant in 1996 was roughly 62 to 1. Theprobability of such an odds ratio occurring by chance is less than one in ten thousand. Fora Hispanic applicant, the odds ratio was approximately 1.19 to 1, and for Asians it wasapproximately 0.77 to 1 (see Figure 5). The Hispanic-white and Asian-white odds ratiosare not statistically significant.

In other words, controlling for the other factors, UMSM likely grants preferencesto black over white applicants. There is, however, no statistical evidence that Hispanics

21

and Asians receive preferences over whites or that whites receive preferences overHispanics or Asians.

1997. The degree of preferences awarded black applicants relative to whites wasless in 1997 than in 1996. Controlling for MCAT scores, science grades, and othervariables, the relative odds of a black applicant being admitted over a white applicantwere roughly 36 to 1. It is less than half of what it was in 1996, but it is still extremelylarge and statistically significant. The Hispanic-to-white odds ratio is 0.57 to 1, which isnot statistically significant. The Asian-to-white odds ratio is 0.38 and is statisticallysignificant,29 raising the possibility of a moderate degree of preference awarded whiteover Asian applicants.

1998. In 1998, racial preferences were also given to black over white applicants,all other things being equal. The odds ratio of black to white applicants that year wasroughly 41 to 1, and is statistically significant. There is no evidence of preferenceawarded Hispanics over whites, as the odds ratio of Hispanics to whites (1.93 to 1) is notstatistically significant. There is some evidence that preferences are given to whites overAsians, however. The Asian-to-white odds ratio in 1998 is 0.62, and is statisticallysignificant. The Asian-to-white odds ratio is only moderate in size.30

1999. In 1999, the degree of preferences awarded blacks relative to whites wasconsiderable, but less than that in 1996, 1997, and 1998. Controlling for MCAT scores,science grades, and other variables, the relative odds of a black applicant being admittedover a white applicant in 1999 were roughly 21 to 1. The odds ratio of Hispanic to whiteapplicants is 2.52, and is not statistically significant. There is no evidence that ethnicpreferences are awarded Hispanic over white applicants. There is also no evidence thatwhites receive preference over Asians in 1999, since the odds ratio of Asian-to-whiteapplicants was 0.68 to 1, which is not statistically significant.

Probabilities of Admission to UMSMThe meaning of the logistic regression equation results and their associated odds

ratios may be difficult to grasp because the equations are complex and hard to explainwithout resorting to mathematical formulations. A more intuitive way to grasp theunderlying dynamic of preferential admissions is to convert these logistic regressionequations into estimates of the probabilities of admission for individuals with differentracial/ethnic group membership, given the same MCAT scores and grades. In this section,we compare the probabilities of admission for individuals belonging to these differentgroups, using the logistic regression equation specific to each year. The probabilitycalculations provide an estimate of the admission chances for members of each group, allwith the same test scores and grades, alumni and residency status, and sex.

We chose to examine the probabilities for an in-state male applicant with noalumni connections to UMSM.31 The same set of test scores and science GPAs areentered for blacks, whites, Hispanics, and Asians. Then we calculated the chances of

29 The reciprocal of the Asian-to-white odds ratio of 0.38 is 2.63 to 1, an odds ratio that is considered to bemoderate in size (see previous discussion on odds ratios and relative strength of association).30 The reciprocal of the Asian-to-white odds ratio of 0.62 is 1.61. This is a moderate association.31 We could have compared probabilities of admission for any combination of alumni status, residencystatus, and sex.

22

admission for a black applicant, a white applicant, a Hispanic applicant, and an Asianapplicant with those academic qualifications. These calculations do not change thestatistical results reported in the earlier section on odds ratios. They simply provide aneasier-to-understand interpretation of their meaning.

The differences in odds ratios translate into large differences in the probability ofadmission based on an applicant’s race. The probability of admission is presented below,first for the 1996 and then for the 1999 applicant pool. We examine the probability ofadmission, keeping constant test scores and grades, and limiting our comparisons to in-state male applicants with no relative having previously graduated from UMSM.

1996. As displayed in Figure 6, a black applicant who was an in-state male, withno relative having graduated from UMSM, with a total MCAT score of 35 and a scienceGPA of 3.00, would have a 36 percent chance of admission. Hispanic, Asian, and whiteapplicants with such academic credentials and similar backgrounds would have had a 1percent chance of admission in 1996.

Figure 6

1996 UMSM Probabilities of Admission

36%

81%

97%100%

1% 1%5%

1%6%

33%

100%

37%

81%

97%

7%

74%

95%

28%

79%

96%

Weighted MCAT=35,Sci-GPA=3.0





Black Hispanic Asian White

23

If the same 1996 black applicant had an MCAT score of 40 and a GPA of 3.25, hewould have had an 81 percent chance of admission, despite a score of 40 being onlyslightly higher than the average for all MCAT test takers in the country.32 That is, havinga score that is only slightly higher than the national average and having roughly a B-plusaverage is still good enough for UMSM to admit eight out of every ten black in-state maleapplicants with no alumni connections. In contrast, an Asian applicant from the samebackground and with the same scores would have had only a 5 percent chance ofadmission, and a similar Hispanic or white applicant, a 7 and 6 percent chance,respectively.

A black applicant with a total MCAT score of 45 and an overall science GPA of3.5 would have had a 97 percent chance of admission in 1996. A Hispanic applicantwould have a 37 percent chance of admission, while a comparable white applicant wouldhave had a 33 percent chance, and an Asian only a 28 percent chance.

With an MCAT score of 50 and a GPA of 3.75, a black applicant was essentiallyguaranteed admission in 1996 (100 percent chance). Similar applicants from the othergroups would also have had a good chance of admission. A Hispanic applicant with thesame credentials and background had an 81 percent chance; a white would have had a 79percent chance; and an Asian, a 74 percent chance.

At the very top, with an MCAT of 55 and a science GPA of 4.0, Hispanic, white,and Asian applicants approach a certainty of admission, controlling for the other factors.With these qualifications in 1996, 97 percent of Hispanics, 96 percent of whites, and 95percent of Asians would be likely to be admitted, as well as all such black applicants.

1999. Figure 7 shows the probabilities of admission for the four groups of in-statemale applicants with no alumni relatives, based on the 1999 applicant pool. Probabilitiesof admission still favor black applicants over those of other groups, but black applicantsare not given quite as much preference relative to other applicants as in 1996.Nevertheless, holding other variables constant, blacks still receive a very large degree ofpreference over other groups.

32 We calculated the national mean to be 37.8. The national writing sample average is a 6, which is thenumerical transformation of the average letter grade of O. The national mean subscores for the verbal,physics, and biology subtests were 8.5, 8.6, and 8.7, respectively. Doubling the writing score and adding itto the remaining mean verbal, physics, and biology subscores gives us a mean weighted total of 37.8 for alltest takers (those rejected as well as those accepted) in 1996. For our studies on medical schools, we have created a total MCAT score, which is the sum of the subtestscores plus a converted score for the writing sample. For the latter, we took the assigned letter grade for anapplicant’s writing sample and converted it into a number. We assigned a 1 to the letter grade of J, a 2 to theletter grade of K, extending to an 11 for a T.

24

Figure 7

1999 UMSM Probabilities of Admission

9%

48%

90%

99%

1% 0%3%0%

4%

31%

100%

53%

92%99%

10%

75%

97%

23%

81%

98%






Black Hispanic Asian White

The gaps are large for those admitted with lower test scores and grades. A blackin-state male applicant with no alumni parents and with a total MCAT score of 35 and acollege GPA of 3.0 would have a 9 percent chance of admission, versus a 1 percentchance for a comparable Hispanic applicant, and no chance for a white or Asian. With anMCAT score of 40 and a GPA of 3.25, a black in-state male applicant with no alumniparents would have a 48 percent chance of admission. A comparable Hispanic applicantwould have a 10 percent chance of admission, while an Asian and a white applicantwould have a 3 and 4 percent chance, respectively.

There are also large gaps between blacks and the other groups for MCAT scoresof 45 and GPAs of 3.5, even when holding constant sex, residency, and whether a studentwas related to a UMSM graduate. Ninety percent of black in-state male applicants withno parents graduating from UMSM would be admitted, as would a small majority (53percent) of comparable Hispanics. A white with the same credentials and backgroundwould have a 31 percent chance of admission, while an Asian with the same academicand nonacademic considerations would have only a 23 percent chance of admission.

With weighted MCAT scores of 50 and science GPAs of 3.75, the differences inadmission probabilities get smaller. With such scores and grades, practically all black in-state male applicants with no alumni parents would be admitted (99 percent), and sowould 92 percent of Hispanics. Whites and Asians would be admitted at somewhat lowerrates (81 percent for whites, 75 percent for Asians).

25

Whites, Hispanics, and Asians reach probabilities identical to the probabilities forblack applicants at the very top. With MCAT scores of 55 and a GPA of 4.00, controllingfor sex, residency, and whether a student was related to a UMSM graduate, theprobability of admission for Hispanics, Asians, and whites—as well as blacks—is closeto 100 percent (99 percent for Hispanics, 97 percent for Asians, 98 percent for whites,and 100 percent for blacks).

26

Consequences of Racial and EthnicPreferences at UMSMEarly Help for URMs

The analysis of MCAT and science GPAs of admittees by race and ethnicityshows that UMSM gives very large preferences in admission to black applicants, even tothe point of accepting many at-risk applicants. Such enrollees enter medical schoolsubstantially underprepared compared to their white and Asian counterparts. Notsurprisingly, UMSM has institutionalized a host of programs designed to prevent these at-risk students from dropping out. There is no indication, however, that UMSM draws adistinction between at-risk and non-at-risk applicants based on test scores andundergraduate grades, rather than race and ethnicity.

UMSM has developed an extensive network of services for URM enrollees tokeep them in school. In its 1996 grant application for federal funding, UMSM defines itsgoal as “develop[ing] a comprehensive academic and non-academic support system forall disadvantaged medical students matriculating at UMSM.” UMSM goes on toexplicitly target blacks for special academic intervention:

This will be achieved by enhancing and expanding current academic developmentefforts for freshman and sophomore students (i.e., first and second-year medicalschool students), with special emphasis on African American students. Therationale for focusing on this group is that the basic science years represent thegreatest hurdle to retention and progression of African-American students.33

The problem of retention and progression of black students, however, is mostly onecreated by the medical school itself. Blacks are admitted with lower test scores and GPAscompared to non-URMs, and the evidence suggests that the medical school knows thatthey are less academically qualified. UMSM’s 1995 Minority Achievement Report claims:

The School of Medicine has continued to struggle with not having enough moneyavailable to make competitive financial aid packages available to the trulycompetitive [URM] applicant whose grades and test scores make them muchsought after and easily wooed by private institutions. . . . This forces the school togive more consideration to the less competitive students whose non-cognitiveattributes become the criteria by which the school looks to give credibility to theirapplication.34

33 University of Maryland School of Medicine, Training Grant Application, Health Careers OpportunityProgram (HCOP Grant Application), Grant No. MB02312-04, 1996, p. 18 (emphasis added).34 Minority Achievement Report, June 1995, pp. 95-96 (emphasis added).

27

Not surprisingly, the medical students are aware of the double standards for someracial and ethnic groups. In its 1996 Minority Achievement Report, UMSM notes,“[Minority students] received unlimited hours for tutoring and other support which isperceived by the non-minority student as ‘special treatment.’”35 UMSM in the report saysnothing about how it might fix this problem.

To keep URMs in school, UMSM offers many types of support to its targetedminorities. Outreach is extensive: Program and support service information andapplications are sent to all targeted minority students admitted to UMSM.36 The medicalschool also makes available a six-week “Pre-Matriculation Summer Program” forminority and other disadvantaged students.37 This is a summer program that goes overcritical portions of the mandatory first-year coursework, even before the in-comingstudents formally take these classes. The program also works on their reading and studyskills, and participants receive counseling even before they start their first year. They mayalso obtain academic tutoring if they anticipate academic difficulty (although tutoring isavailable to others only when they actually face academic difficulty).38

During the academic year, UMSM also provides the following services to URMs:• close monitoring by an academic committee;• Step 1 preparation in the form of study-skills workshops related to Step 1

preparation and test taking, two practice testing sessions for Step 1, reviewmaterials for Step 1, and individual counseling;

• tutoring and practice exams for those receiving D’s or F’s in a course (whoare required to take make-up exams or retake the course),

• minority faculty and student advising;• Stanley Kaplan preparation courses to prepare for testing; and• peer and group tutoring for specific courses, and other interventions.39

UMSM recognizes the need for academic monitoring and intervention from thevery beginning of an at-risk student’s medical education. First, URM students are flaggedby the monitoring committee based on MCAT scores. “The purpose of [UMSM’sacademic monitoring] is to identify and contact first and second-year medical studentswhose test performances indicate a need for academic or non-academic support.”40 Giventhe generally lower test scores of black students, they receive a disproportionate amountof academic monitoring. An average of 25 minority students are discussed at eachacademic monitoring meeting.41

UMSM also provides peer tutoring. In 1994, first-year students received 78 hoursof peer tutoring. Minority freshmen received 56 hours, or 72 percent, of those hours.42

35 Minority Achievement Report, June 1996, p. 87.36 HCOP Grant Application, p. 33.37 Ibid., p. 19.38 Ibid., p. 34.39 The list is drawn from UMSN’s HCOP Grant Application, pp. 55-57. “Step 1” is explained at note 44, infra.40 Ibid., p. 22.41 1996 Minority Achievement Report, p. 89.42 Blacks and Hispanics made up roughly 17 percent of the 1994 freshman class, according to ourcalculations based on computer data obtained by CEO from UMSM.

28

Ninety-five percent of peer tutoring hours for second-year students were received byminority students. Academic counseling hours also went disproportionately towardsminority students. Sixty-nine percent of academic counseling hours for first-year studentswent to minorities, as did 81 percent of hours for second-year students.

Even when hiring tutors, UMSM relies on double standards based on race andethnicity. White and Asian medical students must have had A’s in the courses they are totutor, while minority tutors (blacks, Hispanics, and Native Americans) need only haveB’s, provided they have the requisite but undefined noncognitive factors:

The formal tutorial program employs upper-class students who made “A’s” in thecourse they tutor. . . . Four African American students served as tutors during thepast three years. Non-cognitive factors will be used to select minority tutors whoearned “B’s” in first year courses.43

Analyzing Performance in Medical SchoolCompared to undergraduate education, medical education in America is fairly

standardized. It consists of four years, with basic medical science courses (microbiology,physiology, etc.) making up the first two years, while the third and fourth years involvegreater clinical experience and exposure to a variety of specialties (surgery, obstetrics-gynecology, pediatrics, etc.).

UMSM refused to provide medical school grades and licensing examinationscores to CEO. Instead, CEO obtained information on performance of medical schoolstudents for the 1996 entering class through documents and files unsealed in Farmer v.Ramsay. We have recalculated the statistical information used in the case to match theracial and ethnic group definitions used for this and other CEO reports.

Differences in Medical School GPAs�the First Two YearsDespite extensive intervention by UMSM, blacks had much lower grades on

average during their first two years in medical school compared to Hispanics, Asians, andwhites. These gaps are not surprising, given the much lower MCAT scores of blackscompared to Hispanics, Asians, and whites. Figure 8 shows the distribution of medicalschool GPAs for the first two years.

43 HCOP Grant Application, p. 39.

29

Figure 8

1st & 2nd Year GPAs, UMSM Enrollees(25th, 50th, and 75th percentiles)

2.33

3.00 3.00

2.50

3.00

3.17 3.17

3.00

3.86

3.50

2.67

3.67

2.00

2.50

3.00

3.50

4.00


Black enrollees perform significantly worse than Hispanics, Asians, and whites intheir first two years. The median GPA for blacks is 2.50 at least half a grade-pointlower compared to the others. It is 3.00 for Hispanics and 3.17 for Asians and whites.

Black GPAs for the first two years at the 75th percentile are the same as theHispanic median and the Asian and white GPA at the 25th percentile. This means that 75percent of blacks enrolled at UMSM had lower grades than half the Hispanic enrolleesand 75 percent of Asians and whites. While Asian and white GPAs at the 25th percentileare 3.00, the black GPA at the 25th percentile is 2.33 roughly a C-plus average.

30

Differences in Medical School GPAs�the Third and Fourth Years

Figure 9

3rd & 4th Year GPAs, UMSM Enrollees(25th, 50th, and 75th percentiles)

3.00

3.29

3.44

3.29

3.50 3.503.58

3.47

3.67 3.71

3.31

3.64

2.50

3.00

3.50

4.00


Figure 9 displays the medical school GPAs for the third and fourth years ofmedical school. GPAs are somewhat higher overall than the first two years for all fourgroups, but blacks still lag behind Hispanic, Asian, and white students. The median blackGPA in the third and fourth years is 3.29, compared to 3.50 for Hispanics and Asians, and3.58 for whites. The black median GPA for the third and fourth years is equal to or lowerthan the GPA for the bottom 25th percentile for Hispanics, Asians, and whites. Thismeans that half the black students in their third and fourth years have lower GPAs ascompared to 75 percent or more of Hispanics, Asians, and whites. Black GPAs at the 75th

percentile are also lower than the average GPAs for Hispanics, Asians, and whites.

Differences in Step 1 ScoresDisparities in medical school performance between groups are also reflected in

their performance on Step 1 of the national licensing exam, the United States MedicalLicensing Examination (USMLE).44 These results are also in line with research showing

44 The USMLE is the licensing exam of the National Board of Medical Examiners (NBME). It is generallyrequired of every medical school student seeking to practice medicine in the United States. The USMLEconsists of three separate examinations (Steps 1, 2, and 3). Step 1 is taken after the first two years ofmedical school, and a passing score is often required for a student to continue in medical school. Step 2 istaken during or after the fourth year. Step 3 is taken after graduation from medical school. The NBMEestablishes the minimum scores required to pass each part of the licensing exam. Most scores, according tothe NBME, fall between 160 and 240. The passing score for Step 1 of the USMLE is 179, and the overallpass rate is typically 90 percent. See the USMLE website, <www.usmle.org>.

31

MCATs to be the best predictor of Step 1 scores.45 Since UMSM admitted blacks withsignificantly lower MCAT scores as compared with the other groups, we would expect tofind similar disparities in Step 1 scores.

All medical students at UMSM are required to pass Step 1 before continuing onwith their medical education. A passing score in 1998, when the students whomatriculated in 1996 took the test, was 179. This is the score at the 13th percentile of allStep 1 test takers in 1998 87 percent of those taking the test passed, fewer than 13percent failed. The mean score for first-time test takers in 1998 was 210, and the standarddeviation (the spread around the score that encompasses roughly 68 percent of all testtakers) was 20 points.46

Figure 10

USMLE Step 1 Scores, UMSM Enrollees (25th, 50th, and 75th percentiles)

177

209206

194 195

215

221

213

226

232

184

237

170

180

190

200

210

220

230

240


Figure 10 displays USMLE Step 1 scores at the 25th, 50th, and 75th percentiles forblacks, Hispanics, Asians, and whites taking the test for the first time. The median Step 1score for black students is roughly the same as that for Hispanics, but 11 points lowerthan the Asian median, and 17 points lower than the white median. The Step 1 score forblacks at the 75th percentile is lower than the median score for Asians and whites. It isonly three points higher than the national average for first-time test takers, meaning thatroughly 75 percent of black students at UMSM scored below the national average of allfirst-time test takers, not just those who passed, in 1998.

More significantly, more than a quarter of black students (7 out of 27) who tookthe test in 1998 did not pass Step 1 on the first try, no Asians failed, one Hispanic failed,and two whites (out of 81) failed. The Step 1 score for blacks at the 25th percentile was

45 See, e.g., AAMC, MCAT Interpretive Manual, p. 15. Prediction in performance is further improved,according to the manual, when both types of preadmissions academic qualifications are considered jointly.46 In other words, roughly 68 percent of all test takers had scores falling between 190 and 230.

32

177 two points below the mandatory cut-off score. This puts them at the bottom 13percent of all national test takers. Asian and white students at the 25th percentile did muchbetter. The Asian score at the 25th percentile was 209 and the white score was 206, whichwere only one and four points, respectively, below the national mean. In other words,almost 75 percent of Asian and white UMSM students scored above the national average.

Nine individuals took the test a second time between 1998 and 2000. Sevenpassed. Two had to take Step 1 a third time; one had passed after this third try as ofSummer 2000.

Differences in Step 2 ScoresEighty-four students took Step 2 of the USMLE by July 2000. It is not a

requirement for graduation at UMSM, and so not all students took Step 2 beforegraduating. Figure 11 shows the Step 2 scores by racial and ethnic group. A passing scoreon Step 2 was 170 until May 1, 2000, when it was raised to 174.47 Reporting of nationalpercentiles was discontinued in 1999, so median national scores for Step 2 are notpublicly available.

Figure 11

USMLE Step 2 Scores, UMSM Enrollees(25th, 50th, and 75th percentiles)

206204

196

219214

206

224230

170170

180

190

200

210

220

230

240

Blacks (n=15) Hispanics (n < 5) Asians (n=16) Whites (n=47)

Not Reported

As displayed in Figure 11, most black students who took Step 2 had lower scoresthan most Asians and whites. The median Step 2 score for black students was196 roughly 20 points lower than median Asian and white scores (219 and 214,respectively).

Black scores at the 75th percentile were roughly the same as Asian and whitescores at the 25th percentile. That is, 75 percent of the black students taking Step 2 hadlower scores than roughly 75 percent of Asian and white students.

47 USMLE website, <www.usmle.org>, “Changes in Minimum Passing Score.”

33

About a quarter of black students (4 out of 15) taking Step 2 did not pass; nowhites or Asians failed. The 25th percentile for blacks was 170, which was the passingscore before May 2000, when it was raised to 174.

Relationship between MCATs and Subsequent Performance inMedical School

Disparities in medical school performance at UMSM could have been predictedby the gaps in MCAT scores and science GPAs. Research has shown academicqualifications to be very important in predicting success in medical school. MCAT scoresand undergraduate science grades have been found to be the best predictors of medicalschool performance and passing Step 1 of the USMLE.

The AAMC has been conducting an ongoing study of the validity of the MCATs.In its Interpretive Manual, the AAMC finds the MCATs more valid than other factors inpredicting subsequent performance in medical school.48 The AAMC reports that anindividual’s MCAT scores have a 0.67 correlation with first-year medical school grades, a0.64 correlation with first- and second-year medical school grades, and a 0.72 correlationwith scores on Step 1 of the USMLE. In contrast, an individual’s college science GPAhas a 0.54 correlation with first-year medical school grades, a 0.58 correlation with first-and second-year grades, and a 0.48 correlation with USMLE Step 1 scores.

Other research has also found MCAT scores, more than undergraduate GPAs andany nonacademic traits, to be the best predictor of medical school grades and subsequenttest scores on the medical licensing exams. Wiley and Koenig found MCAT scores to beextremely accurate in predicting first- and second-year medical school grades andUSMLE Step 1 scores. They found the correlation between MCAT scores and USMLEStep 1 scores to be 0.72, and 0.64 for first- and second-year medical school grades,respectively.49 Case, Swanson, Ripkey, Bowles, and Melnick found a statisticallysignificant relationship between MCAT scores and subsequent performance on Step 2 ofthe USMLE, as well as a correlation between MCAT scores and medical students’performance in clinical clerkships.50 In another study, nearly half of all medical studentswith MCAT scores in the bottom quartile of all test takers (a mean score lower than an 8)fail Step 1 on their first try.51 Other researchers have found that matriculants with low 48 Association of American Medical Colleges, MCAT Interpretive Manual: A Guide for Understanding andUsing MCAT Scores in Admissions Decisions (Washington, DC: Association of American MedicalColleges, 1998): 15-16.49 See J.A. Koenig and A. Wiley, “The Validity of the Medical College Admission Test for PredictingPerformance in the First Two Years of Medical School,” Academic Medicine, 71, #10 (October 1996Supplement): S83-S85.50 S.M Case, D. B. Swanson, D.R. Ripkey, L. T. Bowles, and D. E. Melnick, “Performance of the Class of1994 in the New Era of USMLE,” Academic Medicine, 71, #10 (October 1996 Supplement): S91-S93. Seealso K. L. Huff, J.A. Koenig, M. M. Treptau, and S. G. Sireci, “Validity of MCAT Scores for PredictingClerkship Performance of Medical Students Grouped by Sex and Ethnicity,” Academic Medicine, 74, #10(October 1999 Supplement): S41-S44, where a correlation between MCATs and third-year clerkship gradeswas found.51 Roughly 90 percent of all test takers pass Step 1 at any given time. A. Tekian, R. Mrtek, P. Syftestad, R.Foley, and L. J. Sandlow, “Baseline Longitudinal Data of Undergraduate Medical Students at Risk,”Academic Medicine, 71, #10 (October 1996 Supplement): S86-S90; J.A. Koenig, W. Li, and R. Haynes,“Estimation of the Validity of the 1991 MCAT for Predicting Medical School Grades, NBME Performance,

34

MCAT scores—that is, students with mean MCAT scores below 7.0—were at risk foracademic failure, meaning failure to complete medical school.52

We performed similar calculations for UMSM enrollees.53 The correlationbetween MCAT scores and first- and second-year grade point average is 0.66, while thecorrelation between MCAT scores and third- and fourth-year grades is much lower (0.28).

MCATs and USMLE scores are also highly correlated. The correlation betweenMCAT scores and USMLE Step 1 scores on an individual’s first attempt at the test is0.70. The correlation between MCAT scores and their USMLE Step 2 scores is 0.65. Allcorrelations were statistically significant at the 0.0001 level of significance except for thecorrelation between MCATs and the GPA for third- and fourth-year classes, which wasstatistically significant at the 0.001 level.

In terms of a white-black test score gap and possible racial bias regarding theMCATs, research shows little or no racial or ethnic bias associated with the MCATs in itsprediction of subsequent performance. One study found that MCAT scores predictedmedical school performance equally well for all racial and ethnic groups.54 Another studyfound that controlling for MCAT scores and college grades dramatically reduced thedifferences between racial and ethnic groups in passing Step 1 of the USMLE. With thesame MCAT scores and college grades, Hispanic and black men performed about as wellas white men on Step 1. The same was the case for black women as compared with whitewomen with the same academic credentials, and Hispanic women performed only slightlyworse.55

We performed similar analyses on UMSM enrollees by racial and ethnic groups.There is a statistically significant difference in means for the first- and second-year GPAsamong racial and ethnic groups,with no controls for MCATs and undergraduate scienceGPAs. When mean differences in the first- and second-year GPAs are controlled forMCAT scores and college grades, however, there is no performance gap betweenmembers of different racial and ethnic groups. The relationship between race/ethnicityand performance in the first two years of medical school vanishes.

The same is the case regarding Step 1 scores. There is a statistically significantdifference in mean Step 1 scores between groups. The relationship between race/ethnicityand differences in mean Step 1 scores then vanishes, however, when one controls for

and Academic Difficulty,” paper prepared for the MCAT Evaluation Panel Meeting, December 1987,available at <www.aamc.org/stuapps/admiss/mcat/koeni001.htm>.52 K. L. Huff and D. Fang, “When Are Students Most at Risk of Encountering Academic Difficulty? AStudy of the 1992 Matriculants to U.S. Medical Schools,” Academic Medicine, 74, #4 (April 1999): 454-460.53 These correlations, like those reported in the MCAT Interpretive Manual, are corrected for restriction inrange. In other words, because the range of MCAT scores among enrollees is much more limited thanamong all test takers, they should be statistically adjusted to reflect this restriction. See Jacob Cohen andPatricia Cohen, Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences (Hillsdale,NJ: Lawrence Erlbaum, 1975), and AAMC, MCAT Interpretive Manual, p. 15.54 J.A. Koenig, S.G. Sireci, and A. Wiely, “Evaluating the Predictive Validity of MCAT Scores acrossDiverse Applicant Groups,” Academic Medicine, 73, #10 (October 1998): 1095-1106.55 B. Dawson, C. K. Iwamoto, L. P. Ross, R. J. Nungester, D. B. Swanson, and R. L. Volte, “Performanceon the National Board of Medical Examiners Part I Examination by Men and Women of Different Race andEthnicity,” Journal of the American Medical Association, 272, #9 (September 7, 1994): 674-679.

35

MCATs and science GPAs. This means that overall white-black and Asian-black gaps insubsequent medical school performance are a function of prior academic qualifications.Having admitted and enrolled more blacks with lesser qualifications, it follows that thesestudents would perform worse later on.

Differences in Graduation RatesBlack students also graduate at a lower rate than white counterparts, although they

graduate at a higher rate than Asian students. Below are the four-year graduation rates forblack, Hispanic, Asian, and white students entering in Fall 1996.

• 68 percent of blacks• 63 percent of Asians• Hispanics not reported (fewer than 5)• 82 percent of whites

Sixty-eight percent of blacks entering UMSM in Fall 1996 graduated in fouryears. The black graduation rate is lower than the white (68 percent compared to 82percent). It is five points higher than the Asian rate, despite black students having lowerMCAT scores, undergraduate science grades, medical school grades, and USMLE testscores.

The reasons for not graduating vary. Further analysis of those who did notgraduate sheds some light on graduation rates. Of the blacks who did not graduate, all hadfirst- and second-year GPAs of 2.58 or lower. The median first- and second-year GPA fornongraduating black students was 2.17. For those who had third-and fourth-year gradesbut did not graduate, the median GPA was 2.00, and the third- and fourth-year GPA was3.00 at the 75th percentile. As for Step 1 scores, more than half of nongraduating blacksdid not pass the exam at the first attempt. The median Step 1 score for black nongraduateswas 173, which is lower than the required passing score of 179 for Step 1.

In contrast, nongraduating whites and Asians are a split group. Roughly half thenongraduating group of Asians and whites had relatively poor grades and Step 1 scores ascompared to graduating Asians and whites, but the other half, and definitely the topquarter, had notably high grades and Step 1 scores. Among nongraduating whites andAsians, the median first- and second-year GPAs for Asians and for whites was 3.00.Twenty-five percent of nongraduating Asians and of nongraduating whites had GPAsaround 2.50. Twenty-five percent of whites, however, had a first- and second-year GPAof 3.5 or higher, and 25 percent of Asians had a GPA of 4.0 or higher.

Of nongraduating Asians and whites that had third-and fourth-year grades, themedian was 2.12 for Asians, indicating academic difficulty, and 3.38 for whites, whichshows less academic difficulty. At the 75th percentile, however, the GPA for the third andfourth years was 3.67 for Asians and 3.50 for whites, suggesting that this group in the top25 percent of nongraduating Asians and whites was not graduating for reasons other thanacademic difficulty.

The nongraduating whites and Asians also split into two groups on Step 1. Themedian score for nongraduates was 213 for Asians and 232 for whites. The bottom 25th

percentile was 204 for Asians and 194 for whites. These are still higher than the 75th

36

percentile for black nongraduates (193). The top quarter of nongraduates, moreover, hadhigh Step 1 scores (224 for Asians, and 243 for whites).

The statistics on nongraduates suggests that anywhere between a quarter and ahalf of Asian and white students went elsewhere, were in some joint M.D.-Ph.D.program, or did not graduate for reasons other than academic difficulty. In contrast, thereis ample evidence that, for black enrollees, poor academic performance in medical schoolis related to nongraduation in the vast majority of cases.

UMSM implies in various documents that poor academic performance is relatedto low MCATs, and black students, because of their lower MCATs, are more at risk. It isapparent that only substantial intervention by UMSM keeps these academically marginalstudents on track for graduation.

Other Consequences of Racial and Ethnic PreferencesUMSM’s racial and ethnic preference program not only admits blacks who are at

best marginally qualified, it discriminates against better qualified whites, Hispanics, andAsians. This incurs substantial costs to individual applicants denied admission to UMSMon account of the color of their skin.

The state of Maryland has only one publicly funded medical school. Options forin-state rejectees consist of private medical schools, out-of-state medical schools, orforeign medical schools. In the first case, tuition for state medical schools is lower thanthat for private medical schools. As for applying as a nonresident to another state’smedical school, the applicant is at a substantial admissions disadvantage because of hisnonresident status, and must also pay the higher nonresident tuition (if he or she succeedsin getting in). And, as for applying to foreign medical schools, the applicant mustshoulder substantial costs of tuition and travel, plus the added social burden of living in aforeign country. Moreover, foreign medical school graduates are more likely to fail theU.S. licensing exams.56

Consider also the societal costs incurred by such a program of racial preferences.How much does it cost the taxpayers to train at-risk physicians? How much do additionaltutorial and supplemental programs cost taxpayers? Do racial preference programs createresentment and reinforce negative stereotypes? Do they lower standards for physiciantraining? And, ultimately, how do such students perform as doctors?

The long-range impact of racial and ethnic preferences in medical schooladmissions is not known. The medical establishment claims that racial and ethnicpreferences are needed to increase the number of black, Hispanic, and Native Americandoctors, which in turn improves medical care for patients of the same race. But researchin this area is meager, and a review of the literature on minority health-care andphysicians’ race/ethnicity yields contradictory findings. There is also little research onpreferred enrollees’ performance in medical school, performance on licensing exams, andsubsequent physician performance, as measured by various factors, including ratings byfellow physicians.57

56 Barron’s Guide to Medical and Dental Schools (Hauppague, NY: Barron’s Educational Series, Inc.,1997).57 See Satel, PC M.D., for a review of the research. Satel notes that time spent between physician andpatient is probably the most important factor in the doctor-patient relationship, not the race of the physician.

37

In sum, it is quite clear that UMSM goes to great lengths to admit less qualifiedAfrican Americans at the expense of other, better qualified applicants. It is also clear that,having admitted students who are academically at risk, the medical school must thenexpend a disproportionate amount of time, energy, and resources to keep theseunderqualified, at-risk students in school. What UMSM has done is more than arecruitment and outreach campaign. This is a systematic policy to admit, enroll, andgraduate sufficient numbers of some students because of their skin color, at the expenseof other, more academically qualified applicants who happen to lack that skin color.

Most recently, a study found being a minority to be a risk factor in predicting who would be a problemresident, but here, too, insufficient medical knowledge, poor clinical judgment, and insufficient use of timewere the most frequently reported difficulties. (Medical knowledge and clinical judgment are most closelyrelated to test scores, as previously discussed.) See D. C. Yao and S. M. Wright, “National Survey ofInternal Medicine Residency Programs Directors Regarding Problem Residents,” Journal of the AmericanMedical Association, Sept. 6, 2000, available at <www.jama-ama-assn.org>.

38

Appendix Multiple Logistic Regression Equations, 1996-19991996Variable Unstan. Reg. Coef. Odds RatiosBlack 4.1194* 61.5197*Hispanic .1778 1.1946Asian -.2619 .7696Female .6169** 1.8532**In-State Resident 2.9702* 19.4960*Alumni Parent 4.2913* 73.0599*Weighted MCAT Score .2054* 1.2280*Science GPA 3.8902* 48.9227*Constant -26.5243*

*Statistically significant at p < .0001**Statistically significant at p < .01

1997Variable Unstan. Reg. Coef. Odds RatiosBlack 3.5817* 35.9345*Hispanic -.5669 .5673Asian -.9706** .3789**Female .4122# 1.5101#

In-State Resident 3.2172* 24.9592*Alumni Parent 5.2274* 186.3135*Weighted MCAT Score .2289* 1.2573*Science GPA 4.0104* 55.1717*Constant -28.0143*

*Statistically significant at p < .0001** Statistically significant at p < .001# Statistically significant at p < .005

39

1998Variable Unstan. Reg. Coef. Odds RatiosBlack 3.7071* 40.7375*Hispanic .6566 1.9282Asian -.4774## .6204##

Female .5447## 1.7241##

In-State Resident 2.9494* 19.0945*Alumni Parent 3.3932* 29.7615*Weighted MCAT Score .2289* 1.2572*Science GPA 3.8705* 47.9654*Constant -27.5471*

*Statistically significant at p < .0001##Statistically significant at p < .05

1999Variable Unstan. Reg. Coef. Odds RatiosBlack 3.0270* 20.6343*Hispanic .9223 2.5149Asian -.3852 .6803Female .9603* 2.6126*In-State Resident 2.8980* 18.1378*Alumni Parent 2.8724* 17.6791*Weighted MCAT Score .2564* 1.2922*Science GPAs 3.9854* 53.8085*Constant -29.1983

*Statistically significant at p < .0001

CENTER FOR EQUAL OPPORTUNITY

The Center for Equal Opportunity (CEO) is a non-profit research institutionestablished under Section 501(c)(3) of the Internal Revenue Code. CEO sponsors

conferences, supports research, and publishes policy briefs and monographs on issuesrelated to race, ethnicity, immigration, and public policy.

Linda Chavez, President

racial and ethnic preferences and consequences at the … · • black applicants to umsm are given...

Documents